UAI Explorer 2.0 GitHub Feature Survey

Created: 2026-06-20

Purpose

Advance UAI Explorer from a safe documentation portal into a device-aware, version-aware knowledge salvage and lineage system.

The user goal is not only prettier UI. The deeper goal is to prevent useful ideas from disappearing across device drift, stale project versions, old archives, and disconnected machines.

Current UX Gap

The current enterprise shell is functional and safe, but plain:

• It has a flat catalog.
• It has limited project/version hierarchy.
• It has no device survey model.
• It has no drift view.
• It has no idea lifecycle.
• It has no lineage view for "this came from Cadence v5 and may matter later."
• It treats catalog entries as documents, not as living artifacts with origin, status, confidence, and reuse potential.

UAI Explorer 2.0 should feel less like a secure file portal and more like an internal intelligence library.

Feature Clone Targets

1. OpenMetadata

Source: https://github.com/open-metadata/OpenMetadata

Clone-worthy ideas:

• Unified knowledge graph for assets, ownership, classifications, domains, conversations, memories, runbooks, and decisions.
• Human and AI context are first-class, governed surfaces.
• Memory nuggets tied to assets and business context.
• Stewardship and review workflows.

Apply to Explorer:

• Add an artifacts model that can represent files, docs, ideas, decisions, code modules, device folders, and runbooks.
• Add ownership, status, confidence, source device, source project, source version, and sensitivity fields.
• Add "memory nuggets" extracted from old docs or AI sessions.
• Add review states: new, triaged, promoted, merged, parked, superseded, archived.

Why it matters:

OpenMetadata is the closest conceptual match to "trusted governed context for humans and AI", even though it is aimed at data teams.

2. DataHub

Source: https://github.com/datahub-project/datahub

Clone-worthy ideas:

• Data discovery, governance, observability, and lineage in one system.
• Metadata graph that stays fresh through ingestion.
• Human and AI discovery over the same governed catalog.

Apply to Explorer:

• Add a device ingestion pipeline that surveys each machine and records what was seen, when, and from where.
• Add lineage edges:
• derived_from
• supersedes
• forked_from
• candidate_for
• duplicate_of
• belongs_to_version
• Add drift warnings when one device has project material not seen elsewhere.

Why it matters:

Device drift is basically metadata drift. DataHub's core mental model maps well to "what exists, where did it come from, can I trust it, and what changed?"

3. Outline

Source: https://github.com/outline/outline

Clone-worthy ideas:

• Fast team knowledge base.
• Clean document-first UX.
• Collections and nested docs.
• Polished writing and reading experience.

Apply to Explorer:

• Replace the plain card grid with collections:
• Devices
• Projects
• Versions
• Ideas
• Runbooks
• Decisions
• Archives
• Add a stronger document reader with page outline, backlinks, related items, and provenance.
• Add a fast command/search surface.

Why it matters:

Outline is the best model for a polished, calm knowledge workspace.

4. Docmost

Source: https://github.com/docmost/docmost

Clone-worthy ideas:

• Collaborative wiki/documentation UI.
• Notion/Confluence-style page organization.
• Realtime collaboration direction.
• Draw.io/Miro-style visual collaboration topics.

Apply to Explorer:

• Add a space/page hierarchy for project docs.
• Add "canvas" or "map" views later for idea clusters and project lineage.
• Use a Confluence-like admin flow for permissions and spaces, but keep the implementation smaller.

Why it matters:

Docmost points toward a professional documentation workspace rather than a raw admin tool.

5. Logseq

Source: https://github.com/logseq/logseq

Clone-worthy ideas:

• Privacy-first knowledge graph.
• Daily/journal-driven capture.
• Bidirectional linking.
• Multi-device sync and collaboration concerns.
• Strong warning culture around backups during beta sync work.

Apply to Explorer:

• Add daily survey notes per device.
• Add backlinks between ideas, project versions, and decisions.
• Add "capture first, organize later" inboxes for rough ideas.
• Add backup-required warnings before device import or cleanup.

Why it matters:

Logseq fits the user's "good idea never disappears" requirement.

6. SiYuan

Source: https://github.com/siyuan-note/siyuan

Clone-worthy ideas:

• Fine-grained block references.
• Markdown WYSIWYG.
• Block zoom-in.
• Custom attributes.
• Large document editing.

Apply to Explorer:

• Store extracted highlights as addressable blocks.
• Allow an idea from Cadence v5 to be linked without importing the whole stale document into an active project.
• Add custom attributes like reuse_candidate, source_device, staleness, project_version, and review_owner.

Why it matters:

Block-level references are exactly what old-version salvage needs.

7. BookStack

Source: https://www.bookstackapp.com/

Clone-worthy ideas:

• Simple library metaphor: shelves, books, chapters, pages.
• Friendly organization for non-technical users.

Apply to Explorer:

• Use the metaphor, not necessarily the exact structure:
• Shelf = domain or device group.
• Book = project.
• Chapter = version or category.
• Page = document or artifact.

Why it matters:

Explorer needs a mental model that a tired human can understand quickly.

8. Wiki.js

Source: https://js.wiki/

Clone-worthy ideas:

• Groups, permissions, and page rules.
• Path-based permission rules.
• Theme and dark-mode direction.

Apply to Explorer:

• Extend current tags into saved access groups.
• Add admin-visible access previews:
• "This user can see 42 artifacts."
• "This tag grants access to these projects."
• "This device import is hidden from public-safe viewers."

Why it matters:

The current tag system works, but the admin UX needs better permission explainability.

9. FormKiQ

Source: https://github.com/formkiq/formkiq-core

Clone-worthy ideas:

• Document ingestion.
• Metadata tagging.
• Full-text and OCR search.
• Versioning, retention, and audit trails.
• Event hooks for enrichment and classification.

Apply to Explorer:

• Add ingestion jobs for each device survey.
• Add future OCR/PDF/image text extraction.
• Add retention states:
• active
• reference
• frozen
• archive
• delete_candidate
• Add enrichment hooks for AI summary, tag proposal, duplicate detection, and secret scan.

Why it matters:

This is the document-management side of Explorer 2.0.

10. ArchiveBox

Source: https://github.com/archivebox/archivebox

Clone-worthy ideas:

• Preserve snapshots as ordinary files plus metadata.
• Multiple capture formats.
• Useful archives, not proprietary silos.

Apply to Explorer:

• Before cleanup, snapshot old project folders into a content-addressed archive.
• Store survey metadata in SQLite/JSON while keeping original files readable.
• Support "cold archive, warm index": old material is safe, searchable, but not confused with active project truth.

Why it matters:

Device cleanup needs a preservation layer before any deletion or consolidation.

11. AFFiNE

Source: https://github.com/toeverything/AFFiNE

Clone-worthy ideas:

• Local-first workspace framing.
• Docs, canvas, and tables as related ways to inspect the same knowledge.
• Blocks and object-style organization without forcing a rigid wiki tree.

Apply to Explorer:

• Borrow the multi-mode idea, but keep it smaller:
• catalog cards,
• lineage rail,
• idea lanes,
• optional canvas/map later.
• Do not build a full whiteboard until relationships are real.

12. AppFlowy

Source: https://github.com/AppFlowy-IO/AppFlowy

Clone-worthy ideas:

• Projects, wikis, databases, templates, and AI in one self-hostable workspace.
• Strong ownership and data-control positioning.
• Kanban/database views for task-like material.

Apply to Explorer:

• Use database-like tables and compact kanban lanes for ideas and review tasks.
• Borrow templates for START_HERE_AI.md, survey reports, and project handoffs.
• Avoid becoming a general Notion clone.

13. Anytype

Source: https://github.com/anyproto/anytype-ts

Clone-worthy ideas:

• Offline-first, peer-to-peer, encrypted knowledge OS.
• Custom types and relations.
• Pages, tasks, wikis, journals, and apps sharing one object model.

Apply to Explorer:

• Treat devices, projects, versions, ideas, runbooks, and artifacts as typed objects with relations.
• Keep public Explorer as a sanitized output surface; raw/local-first sync belongs in the survey/redaction zone.

14. Papermerge

Source: https://github.com/ciur/papermerge

Clone-worthy ideas:

• OCR-backed digital archive.
• Tags, hierarchical folders, full-text search, and desktop-like browser.
• Long-term scanned/document archive focus.

Apply to Explorer:

• Add OCR/text extraction later for PDFs, images, scans, and old exported docs.
• Preserve long-term archives with metadata and searchable sanitized text.

15. Onyx

Source: https://github.com/onyx-dot-app/onyx

Clone-worthy ideas:

• Hybrid vector plus keyword search.
• Background workers for connector sync.
• Document sets, role-based access, audit/query history, and air-gapped deployment options.

Apply to Explorer:

• Add safe embedding lanes after air-gapped redaction is in place.
• Keep retrieval scoped by catalog tags, access groups, and sanitized document sets.

16. Open WebUI Knowledge

Source: https://docs.openwebui.com/features/workspace/knowledge/

Clone-worthy ideas:

• Knowledge collections scoped to models.
• Focused retrieval versus full-context mode.
• Nested directories and incremental directory sync.

Apply to Explorer:

• Let AI contexts choose between:
• summary packet,
• focused retrieval,
• full approved document context.
• Never let full-context mode bypass redaction or tag authorization.

17. Graphiti

Source: https://github.com/getzep/graphiti

Clone-worthy ideas:

• Temporal context graphs for AI agents.
• Facts with validity windows and provenance.
• Querying across time, meaning, and relationships.

Apply to Explorer:

• Model project truth over time:
• "Cadence v5 said X",
• "Cadence v6 superseded X",
• "X may return in v7."
• Use temporal provenance before a visual graph.

18. LightRAG

Source: https://github.com/HKUDS/LightRAG

Clone-worthy ideas:

• Graph plus vector retrieval.
• Incremental updates for large knowledge sets.
• Better global context than flat chunk search.

Apply to Explorer:

• Prototype a graph/vector lane only after sanitized artifacts and lineage edges exist.

19. OpenViking

Source: https://github.com/volcengine/OpenViking

Clone-worthy ideas:

• Agent context as a filesystem-like hierarchy.
• Tiered context loading.
• Recursive directory retrieval plus semantic search.

Apply to Explorer:

• Keep START_HERE_AI.md, project summaries, runbooks, and deep references in clear L0/L1/L2 context tiers.
• This is the best pattern for AI worker UX.

20. Dify

Source: https://github.com/langgenius/dify

Clone-worthy ideas:

• Knowledge pipeline from ingestion through retrieval.
• Visual workflow thinking.
• RAG pipeline, prompt IDE, observability, and API-first app building.

Apply to Explorer:

• Use a visible pipeline model:
• survey,
• redact,
• curate,
• publish.
• Keep this as operator workflow status, not a visual workflow builder.
• Add prompt/context template management later for AI workers.

21. AnythingLLM

Source: https://github.com/Mintplex-Labs/anything-llm

Clone-worthy ideas:

• Workspaces as context boundaries.
• Drag-and-drop document ingestion.
• Source citations and workspace-local agents.

Apply to Explorer:

• Treat projects or access groups as context workspaces.
• Add source citations and evidence links to AI context packets.
• Do not let workspace convenience bypass tag or redaction rules.

21a. DocsGPT

Source: https://github.com/arc53/DocsGPT

Clone-worthy ideas:

• Chat/search over documentation.
• Source-focused retrieval.
• Simple user experience around asking documents questions.

Apply to Explorer:

• Use question-oriented discovery later, but only over approved catalog entries and sanitized context tiers.
• Keep the current reader first; chat should explain documents, not replace catalog governance.

22. Khoj

Source: https://github.com/khoj-ai/khoj

Clone-worthy ideas:

• Personal AI second brain.
• Search across notes, documents, and web/context sources.
• Scales from local personal use to larger deployments.

Apply to Explorer:

• Use "second brain" behavior for private lab memory, but expose only sanitized catalog output publicly.
• Add saved searches and question-oriented discovery after device surveys.

23. TriliumNext

Source: https://github.com/TriliumNext/trilium

Clone-worthy ideas:

• Deep note trees.
• Notes can appear in multiple tree locations.
• Rich notes, code notes, relations, maps, import/export, and revision history.

Apply to Explorer:

• Let one artifact belong to multiple views:
• device,
• project,
• version,
• idea lane,
• access group.
• Use this to avoid losing stale ideas simply because they do not fit one current project.

24. RAG-Anything

Source: https://github.com/HKUDS/RAG-Anything

Clone-worthy ideas:

• Multimodal RAG over documents with text, tables, equations, and images.
• Cached parsing results.
• External parser integration.

Apply to Explorer:

• Later ingestion should preserve parsed artifacts separately from raw files.
• Store sanitized parse results and hashes for PDFs, screenshots, diagrams, and exported docs.

25. Foam

Source: https://github.com/foambubble/foam

Clone-worthy ideas:

• Markdown knowledge management on top of VS Code and GitHub.
• Wikilinks, backlinks, graph visualization, daily notes, and publishing.
• Strong ownership of plain text knowledge.

Apply to Explorer:

• Add backlinks and relationship edges before a full visual graph.
• Keep survey notes and handoffs plain-text/exportable.
• Use Git-like provenance ideas for device drift and stale material.

26. SilverBullet

Source: https://github.com/silverbulletmd/silverbullet

Clone-worthy ideas:

• Programmable, private, browser-based personal knowledge database.
• Wiki-style linking, built-in database/query language, and scripting.
• Notes as a programmable system rather than static pages.

Apply to Explorer:

• Treat review queues, idea lanes, and context tiers as queryable views over artifacts.
• Keep programmable/queryable behavior internal; public users should see curated outputs, not arbitrary scripts.

27. Backstage

Source: https://github.com/backstage/backstage

Clone-worthy ideas:

• Centralized developer portal.
• Software catalog.
• Templates for standard practices.
• TechDocs with a docs-like-code approach.

Apply to Explorer:

• Build a UAI documentation spine: project, source version, standards, catalog entry, reader, AI packet.
• Treat Devine/Cadence standards like governed developer-portal assets, not loose files.
• Add ownership, source device, and freshness to every standards page.

28. Docusaurus

Source: https://github.com/facebook/docusaurus

Clone-worthy ideas:

• Easy-to-maintain open-source documentation websites.
• Version-aware documentation.
• Sidebar/navigation patterns that make docs feel structured.

Apply to Explorer:

• Treat Cadence v5/v6/v7 and Devine standards as versioned documentation sets.
• Add reader-side version and lineage affordances before a complex graph.

29. Material For MkDocs

Source: https://github.com/squidfunk/mkdocs-material

Clone-worthy ideas:

• Professional Markdown docs with strong navigation and search ergonomics.
• Works well across devices.
• Plain Markdown authoring with polished output.

Apply to Explorer:

• Keep the human reader calm, searchable, and responsive.
• Borrow the "docs that simply work" standard for runbooks, standards pages, and START_HERE files.

30. MarkItDown

Source: https://github.com/microsoft/markitdown

Clone-worthy ideas:

• Converts PDFs, Office files, images, audio, HTML, CSV, JSON, ZIP contents, and other formats to Markdown for LLM/text-analysis pipelines.
• Explicitly warns that conversion runs with current process privileges.

Apply to Explorer:

• Use conversion only inside the raw survey/redaction zone.
• Convert MacBook Devine/Cadence exports to Markdown, then redact and review before catalog publication.
• Store conversion provenance and sanitized output hashes.

31. Plane

Source: https://github.com/makeplane/plane

Clone-worthy ideas:

• Work items, cycles, modules, saved views, pages, and analytics.
• Turns loose ideas into actionable work without becoming only a document app.

Apply to Explorer:

• Make review tasks first-class: survey, classify, redact, promote, publish.
• Add saved views later for "Mac survey", "Cadence salvage", and "redaction queue".

32. Huly

Source: https://github.com/hcengineering/platform

Clone-worthy ideas:

• All-in-one platform spanning project management, chat, CRM, HRM, and ATS.
• A typed platform model for business applications.

Apply to Explorer:

• Borrow the typed-platform idea, not the breadth.
• Keep devices, standards, ideas, surveys, and artifacts as typed objects with focused workflows.

33. Memos

Source: https://github.com/usememos/memos

Clone-worthy ideas:

• Quick capture.
• Markdown-native portable notes.
• Timeline-first simplicity.

Apply to Explorer:

• Add lightweight capture lanes for ideas discovered during device surveys.
• Avoid forcing every rough idea into a project tree immediately.

34. Dendron

Source: https://github.com/dendronhq/dendron-site

Clone-worthy ideas:

• Local-first Markdown knowledge management for developers.
• Gradual structure as the knowledge base grows.
• Lookup, schemas, backlinks, note references, and graph navigation.

Apply to Explorer:

• Use schemas/templates for START_HERE, project brief, standard, survey, and salvage notes.
• Let MacBook material start rough and gain structure through review.

35. Pagefind

Source: https://github.com/Pagefind/pagefind

Clone-worthy ideas:

• Static low-bandwidth search at scale.
• No hosted infrastructure requirement.

Apply to Explorer:

• Prototype local sanitized search before full embeddings.
• Keep public search scoped to sanitized catalog output.

36. DocSearch

Source: https://github.com/algolia/docsearch

Clone-worthy ideas:

• Crawls documentation into an index and provides a dropdown search experience.
• Search UI is a product feature, not just a backend endpoint.

Apply to Explorer:

• Add a command/search surface that previews project, device, version, and standards context.
• Prefer Pagefind or self-hosted Typesense-style search for private/internal deployments where external indexing is not acceptable.

37. La Suite Docs

Source: https://github.com/suitenumerique/docs

Clone-worthy ideas:

• Clean structured documents.
• Real-time collaboration.
• Granular access control.
• Import/export across .docx, .md, .odt, and .pdf.
• Public-organization posture around ownership and self-hosting.

Apply to Explorer:

• Keep the public-safe reader polished and access-explicit.
• Treat imports/exports as governed transitions, not casual file movement.
• Defer real-time editing, but borrow the clean document hierarchy and access language.

38. Qeta For Backstage

Source: https://github.com/drodil/backstage-plugin-qeta

Clone-worthy ideas:

• Internal Q&A connected to a developer portal.
• Questions, answers, articles, and knowledge sharing in one place.
• Search, catalog, notification, and permission integrations.
• AI-powered answers as an optional layer.

Apply to Explorer:

• Add a later "questions answered by docs" lane for operators and AI workers.
• Link questions to projects, versions, devices, and standards.
• Keep Q&A answers grounded in approved catalog entries and redacted packets.

39. nb

Source: https://github.com/xwmx/nb

Clone-worthy ideas:

• Plain-text note taking, bookmarking, archiving, tagging, filtering, search, Git versioning, syncing, and Pandoc conversion in a portable workflow.

Apply to Explorer:

• Keep survey output portable and plain-text friendly.
• Use Git-like history concepts for device survey manifests and standards import notes.
• Prefer boring export formats for lab memory over proprietary capture.

40. Notesium

Source: https://github.com/alonswartz/notesium

Clone-worthy ideas:

• Fast search with note previews and line highlighting.
• Bi-directional links.
• Local web app with tabs, panels, labels, detailed lists, graph panel, and incoming/outgoing link sidebars.
• Relationship graph that can emphasize active notes and filtered matches.

Apply to Explorer:

• Add compact search previews before a full chat layer.
• Add a reader relationship rail with incoming/outgoing links.
• Use graph-lite panels until the lineage model has enough real data for an interactive graph.

Explorer 2.0 Product Model

Recommended top-level entities:

• devices
• surveys
• projects
• project_versions
• artifacts
• artifact_blocks
• ideas
• lineage_edges
• collections
• access_groups
• review_tasks
• import_jobs
• context_workspaces
• context_tiers
• sanitized_exports
• relationship_views
• standards
• documentation_spines
• saved_views
• operator_actions

Recommended statuses:

• new
• active
• needs_review
• stale
• stale_but_valuable
• promoted
• merged
• superseded
• parked
• archived
• delete_candidate

Recommended categories:

• runbook
• diary
• decision
• architecture
• code
• prompt
• idea
• survey
• credential_pointer
• archive
• unknown

Recommended lineage edges:

• derived_from
• supersedes
• forked_from
• duplicate_of
• candidate_for
• belongs_to_version
• observed_on_device
• imported_from
• redacted_from
• references

UX Upgrade Directions

A. Home Dashboard

Replace the plain catalog landing with an operational dashboard:

• Device survey status.
• Recently changed projects.
• Drift warnings.
• Ideas waiting for classification.
• Documents needing review.
• Migration/readiness health.

B. Device Map

Add a device-first view:

• Device cards for NUC, MacBook, dormant machine, lab hosts, and future nodes.
• Last seen.
• Last surveyed.
• Trust level.
• Active project roots.
• Unknown/stale folders count.
• Survey action history.

C. Project Timeline

Add a project/version view:

• Cadence v5, v6, v7 candidate.
• Which devices contain each version.
• What was superseded.
• What may be reusable.
• What is canonical today.

D. Idea Inbox

Add an idea salvage lane:

• New capture.
• Needs classification.
• Parked for later.
• Promoted to project.
• Merged into current docs.
• Superseded but retained.

E. Lineage Graph

Add a graph or relationship panel before a full visual graph:

• "This came from..."
• "This replaced..."
• "This might belong in..."
• "Similar artifacts..."
• "Referenced by..."

The first implementation should be a relationship map with readable edges, not a canvas graph. Borrow Foam/SilverBullet backlink thinking and DataHub lineage semantics, but keep it scan-friendly.

F. Review Queue

Add an operator workflow:

• Secret scan needed.
• Duplicate suspected.
• Old version needs triage.
• Missing START_HERE file.
• Device survey stale.
• Access group review needed.

The dashboard should show this queue explicitly so Explorer does not become a passive library.

G. Documentation Spine

Add a narrow, visible path from source to safe output:

• catalog,
• version,
• conversion,
• search,
• publish.

This borrows from Backstage TechDocs, Docusaurus, Material for MkDocs, MarkItDown, Pagefind, and DocSearch. It keeps the product feeling professional without becoming a full developer portal.

H. Design Standards Intake

Add a standards-focused intake view:

• Devine UI doctrine,
• Cadence v5 UX fragments,
• Cadence v6 active standards,
• UAI Explorer UI rules.

Each standard should show source device, freshness, review state, tags, and publication treatment.

I. Context Stack

Add a context-tier model before embedding search:

• L0 Start Here: current truth and next command.
• L1 Project Brief: canonical project summary.
• L2 Runbooks: approved procedures and evidence.
• L3 Retrieved Context: scoped sanitized retrieval.
• L4 Archive References: stale-but-valuable material.
• R Raw Survey Zone: restricted, not mounted publicly.

This borrows from OpenViking tiered context, Dify/AnythingLLM knowledge workspaces, and DocsGPT-style document Q&A while keeping Explorer public-safe.

J. Operator Console

Add a compact start surface for the current work wave:

• saved views,
• next actions,
• redaction posture,
• survey focus,
• standards intake focus,
• context readiness.

This borrows lightly from Plane saved views, Outline/Docmost workspace polish, Qeta knowledge workflows, and local-first capture tools like Logseq, nb, and Notesium. It should stay small: a launch surface, not a project-management app.

K. Command Surface

Add a safe command/search surface:

• dashboard lanes,
• authorized catalog entries,
• standards intake,
• redaction loop,
• AI context tiers,
• operator review queue.

This borrows from DocSearch/Pagefind search ergonomics and Backstage-style portal discovery, but it must only search sanitized destinations already authorized for the current session. It is not raw file search.

L. Collection Shelf

Add a human-readable documentation organization surface:

• shelf,
• book,
• chapter,
• page,
• source family,
• publish state,
• required tags,
• safe destination.

This borrows from Outline collections and nested documents, BookStack shelves, books, chapters, and pages, and Docusaurus sidebars and versioned docs. UAI Explorer should keep the metaphor pragmatic: shelves organize approved or pending knowledge, but raw survey material never becomes a public shelf until sanitized output exists.

M. Source Passport

Add a compact source profile surface:

• owner,
• freshness,
• trust state,
• lineage notes,
• access boundary,
• next action,
• source tags.

This borrows from DataHub/OpenMetadata asset profiles, OpenMetadata ownership and freshness semantics, and Graphiti temporal provenance. UAI Explorer should keep it small: a readable source passport for device and documentation lanes, not a full metadata-platform clone.

N. Survey Intake Workbench

Add a metadata-only intake workbench:

• source device and host,
• survey count versus candidates seen,
• suppressed aliases,
• topic clusters,
• file type mix,
• raw-zone boundary,
• redaction/classification state,
• catalog promotion state.

This borrows from Papermerge/FormKiQ ingestion queues, OpenMetadata/DataHub classification and governance patterns, and Plane/Huly work queues. UAI Explorer should keep this intentionally small: the workbench shows what was observed and what needs review, but it does not expose raw contents or make promotion automatic.

O. Access Layer Preview

Add a permission explainability surface:

• user role,
• required tags,
• human versus AI surface,
• visible material,
• blocked material,
• audit event type,
• hard deny gates.

This borrows from Wiki.js and BookStack permission clarity, La Suite Docs' access-explicit document posture, and Outline/Docmost collection-style workspace language. UAI Explorer should not become a full IAM product here; the useful clone is the preview: before giving someone a login, the operator can see what a Cortex tag, code tag, AI role, or admin role actually changes.

P. Mode-Governed Navigation

Separate global navigation by user mode:

• human work,
• AI surface,
• admin ops.

This makes the human/AI divide visible at the shell level and keeps admin-only surfaces from feeling like ordinary reader navigation.

Q. Better Visual Design

The current shell is restrained and safe, but plain. Upgrade with:

• Denser dashboard summary bands.
• Left sidebar grouped by work mode.
• Safe command/search surface.
• Collection shelf for project, standards, archive, runbook, and restricted raw-zone organization.
• Compact public-safe trust band in place of a marketing-style hero.
• Secondary right rail for lineage/context.
• Project/version badges with consistent color semantics.
• Timeline strips for versions.
• Device health chips.
• Confidence/status badges.
• Source passports with owner, freshness, trust, lineage, and access boundary.
• Survey intake workbench for metadata clusters waiting on redaction.
• Access layer preview for roles, tags, blocked surfaces, and audit state.
• Documentation spine.
• Standards intake matrix.
• Empty states with direct actions.

Avoid:

• Decorative dashboards with no workflow value.
• Big marketing hero sections.
• A one-note color palette.
• Graph visuals before the underlying model is useful.

First Implementation Slice

Build Explorer 2.0 in this order:

1. Add data model for devices, surveys, artifacts, and lineage_edges.
2. Add device survey runbook and START_HERE_AI.md template.
3. Add a dashboard view.
4. Add a device inventory view.
5. Add project/version fields to catalog items.
6. Add an idea inbox.
7. Add lineage panel on reader pages.
8. Add import/review queues.
9. Add survey intake for raw-zone metadata clusters.
10. Add collection shelf for human-readable documentation organization.
11. Add access layer preview for role/tag permission explainability.
12. Add standards intake and documentation spine.
13. Add operator console saved views and next-action jump targets.
14. Add full graph view only after relationships are real.

Best Clone Targets By Priority

1. OpenMetadata for governed context and memory.
2. DataHub for lineage and drift thinking.
3. Outline for polish and document UX.
4. Logseq/SiYuan for backlinks and block-level idea salvage.
5. ArchiveBox for preservation-before-cleanup.
6. BookStack/Wiki.js for approachable hierarchy and permissions UX.
7. Backstage/TechDocs for a documentation spine and developer-portal posture.
8. Docusaurus/Material for MkDocs for versioned docs and polished reader UX.
9. FormKiQ/MarkItDown for ingestion, metadata, retention, and conversion.
10. Plane/Huly for review-task workflows and saved views.
11. Memos/Dendron for quick capture, gradual structure, and developer PKM.
12. Anytype/AFFiNE/AppFlowy for typed workspace objects, lanes, and templates.
13. Papermerge for long-term document archive and OCR posture.
14. Pagefind/DocSearch for fast sanitized documentation search.
15. La Suite Docs/Qeta/nb/Notesium for access-explicit docs, knowledge Q&A, portable local notes, and relationship previews.
16. Dify/AnythingLLM/Onyx/Open WebUI for scoped AI knowledge sets after redaction is air-gapped.
17. Khoj/TriliumNext for second-brain search, deep trees, and multi-parent organization.
18. Graphiti/LightRAG/OpenViking for temporal agent context and tiered memory.
19. RAG-Anything for multimodal parse results after safe ingestion exists.

What To Clone Now Versus Later

Clone now:

• Dashboard operating loop.
• Device survey status.
• Idea lanes.
• Review queue.
• Context stack.
• Relationship map.
• Reader context rail.
• Redaction boundary indicators.
• Collection shelf.
• Survey intake workbench.
• Access layer preview.
• Feature radar and review queue.
• Documentation spine.
• Standards intake matrix.
• Operator console with saved views and next actions.

Clone after the backend model exists:

• Device survey import jobs.
• Typed artifact relations.
• Lineage edges.
• Review task tables.
• Safe summary/embedding lanes.
• Context workspace and tier metadata.
• Relationship map persisted from real lineage edges.
• Persisted standards and saved views.
• Sanitized Pagefind or Typesense search index.

Clone later or avoid:

• Full whiteboard canvas.
• Heavy realtime collaboration.
• Full Notion replacement features.
• Raw whole-device embedding.
• Graph visualizations before graph data exists.

Design Principle

Explorer 2.0 should not ask, "Which file is current?"

It should answer:

• What exists?
• Where did it come from?
• Which device saw it?
• Which version does it belong to?
• Is it safe to show?
• Is it stale, canonical, or reusable later?
• What should a human or AI do next?